Vous êtes ici : Accueil » Kiosque » Annonce


Mot de passe : 

Mot de passe oublié ?
Détails d'identification oubliés ?


1 septembre 2017

Phd position: “Complex scene analysis with non-conventional cameras for robotics and people assistance” Le2i, University of Burgundy / RoPeRT, University of Zaragoza

Catégorie : Doctorant

Une bourse de thèse en co-tutelle internationale entre le Le2i (Université de Bourgogne Franche-Comté) et RoPeRT (Université de Saragosse) sur l'analyse de scènes à partir d'imageurs non-conventionnels est à pourvoir.

Début du contrat : Octobre-Novembre 2017.


Phd offer: “Complex scene analysis with non-conventional cameras for robotics and people assistance


Keywords: Computer vision, machine learning, optimization methods

Nowadays, we can extract numerous data on a scene thanks to different kind of cameras. A RGB camera can acquire the photometric information, a RGB-D camera can easily give us the deep, an omnidirectional camera can provide a panoramic view in one shot, thermal vision or multi-spectral cameras can also offer different modalities. All these data provide us important information on the scene to enrich and to make evolve the perception of the environment in which these sensors are moving.

In a static scene, we have shown that we can precisely localize a robot thanks to: a pinhole camera with prior knowledge about the 3D structure of the scene [1] or knowledge about the motion [2,3]; using a RGBD camera [4]; a combination of omnidirectional camera and RGBD camera [5]. Unfortunately, these methods are not valid when the scene is strongly dynamic. Thus, these methods have to be adapted to deal with more complex scenes.

In this work, we will investigate techniques of computer vision, robotics and machine learning to combine these different sensors (RGBD, omnidirectional camera, thermal imaging camera…) by exploiting the amount of data generated and the benefits from its egocentric point of view and different modalities in a life-long concept for enhancing the perception of the environment.

This PhD offer is an international position where the work will be conjointly done between University of Zaragoza, Spain (Robotics, Perception and Real Time group) and University of Burgundy, France (Le2i laboratory, Vision for Robotics group). The applicant will receive the both degrees from the two universities.

About the research groups:

The RoPeRT group is one of the top research groups in robotics and computer vision in Spain, it is renowned nationally and internationally. The group offers a unique opportunity for professional development of a doctoral student. The research team in the project have an extensive record of publications in major journals (TRO, PAMI, IJCV ...) and the most prestigious forums (ICCV,ECCV,ICRA,IROS ...) in robotics and computer vision. Also, the group has connections with top research groups internationally, both academic and professional, for the establishment of partnerships, research visits and internships.

The Vision for Robotics group of the Le2i is working on computer vision methods for robotics application. Its research activity combines non-classical imaging and computer vision (physical and geometrical vision) for enhancing the robot perception. It also maintains strong activity in technology transfer, as evidenced by national and international patents, by incubation of companies and by leading several contracts for applied research. The members of the team are also involved in many international programs for research and education (Erasmus Mundus Masters in Vision and Robotics, Eureka projects, international partnerships…).

About the candidates:

We are looking for a highly motivated person with a M.Sc. in Computer Science/Engineering, Electrical/Electronics/Automation Engineering, Mathematics or similar.

Applicants will be ranked based on: academic record, experience in robotics, machine learning and computer vision, good programming skills (C++, Matlab and/or Python) and good background in statistics and linear algebra. Applicants should have a working English level or higher.

Deadline for applications: September 30th, 2017
Starting date: Between October and November 2017

Candidacy: please send ASAP an email with your resume, academic record, recommendation letter(s) to:

J.J. Guerrero josechu.guerrero@unizar.es
Cédric Demonceaux cedric.demonceaux@u-bourgogne.fr

[1] “Robust and Optimal SoS-based Point-to-Plane Registration of Image Sets and Structured Scenes “Danda Pani Paudel, Cédric Demonceaux, Adlane Habed, Pascal Vasseur. In IEEE/CVF International Conference on Computer Vision (ICCV 2015), December 2015, Santiago, Chile.

[2] “Homography Based Egomotion Estimation with a Common Direction.” O. Saurer, P. Vasseur, R. Boutteau, C. Demonceaux, M. Pollefeys, F. Fraundorfer. In IEEE Transaction on Pattern Analysis and Machine Intelligence, Vol. 39, Issue 2, February 2017, pp 327-341.

[3] “True Scaled 6 DoF Egocentric Localisation with Monocular Wearable Systems”, D. Gutierrez-Gomez, J.J. Guerrero, In Image and Vision Computing, 52, 2016, pp:178-194

[4] "Inverse Depth for Accurate Photometric and Geometric Error Minimisation in RGB-D Dense Visual Odometry", D. Gutierrez-Gomez, W. Mayol-Cuevas, J.J. Guerrero., In IEEE International Conference on Robotics and Automation (ICRA), Seattle May 2015

[5] “Peripheral Expansion of Depth Information via Layout Estimation with Fisheye Camera”, A. Perez-Yus, G. Lopez-Nicolas, J.J. Guerrero, In 14th European Conference on Computer Vision ECCV2016, Amsterdam, Oct 2016


Dans cette rubrique

(c) GdR 720 ISIS - CNRS - 2011-2015.